Retrospective checking of compliance with practice guidelines for acute stroke care: a novel experiment using openEHR's Guideline Definition Language.

Nadim Anani,Sabine Koch,Rong Chen,Tiago Prazeres Moreira

doi:10.1186/1472-6947-14-39

Abstract

BackgroundProviding scalable clinical decision support (CDS) across institutions that use different electronic health record (EHR) systems has been a challenge for medical informatics researchers. The lack of commonly shared EHR models and terminology bindings has been recognised as a major barrier to sharing CDS content among different organisations. The openEHR Guideline Definition Language (GDL) expresses CDS content based on openEHR archetypes and can support any clinical terminologies or natural languages. Our aim was to explore in an experimental setting the practicability of GDL and its underlying archetype formalism. A further aim was to report on the artefacts produced by this new technological approach in this particular experiment. We modelled and automatically executed compliance checking rules from clinical practice guidelines for acute stroke care.MethodsWe extracted rules from the European clinical practice guidelines as well as from treatment contraindications for acute stroke care and represented them using GDL. Then we executed the rules retrospectively on 49 mock patient cases to check the cases’ compliance with the guidelines, and manually validated the execution results. We used openEHR archetypes, GDL rules, the openEHR reference information model, reference terminologies and the Data Archetype Definition Language. We utilised the open-sourced GDL Editor for authoring GDL rules, the international archetype repository for reusing archetypes, the open-sourced Ocean Archetype Editor for authoring or modifying archetypes and the CDS Workbench for executing GDL rules on patient data.ResultsWe successfully represented clinical rules about 14 out of 19 contraindications for thrombolysis and other aspects of acute stroke care with 80 GDL rules. These rules are based on 14 reused international archetypes (one of which was modified), 2 newly created archetypes and 51 terminology bindings (to three terminologies). Our manual compliance checks for 49 mock patients were a complete match versus the automated compliance results.ConclusionsShareable guideline knowledge for use in automated retrospective checking of guideline compliance may be achievable using GDL. Whether the same GDL rules can be used for at-the-point-of-care CDS remains unknown.

Highlights

Providing scalable clinical decision support (CDS) across institutions that use different electronic health record (EHR) systems has been a challenge for medical informatics researchers
We explore the computerised retrospective checking of compliance with clinical practice guidelines for acute stroke care through executing compliance checking rules written in Guideline Definition Language (GDL)
Through the example of clinical practice guidelines for acute stroke management, we explored the use of GDL in combination with openEHR to perform automatic guideline compliance checking in an experimental setting

Summary

Introduction

Providing scalable clinical decision support (CDS) across institutions that use different electronic health record (EHR) systems has been a challenge for medical informatics researchers. Guidelines may improve care by providing better clinical outcomes, ensuring patient safety, reducing costs and decreasing care variability [1] This has led to a great interest in utilising guidelines for providing patient-specific recommendations at the point of clinical decision making and for checking compliance with guidelines retrospectively. The main approach to achieving computerised execution of guidelines in medical informatics research has been the use of languages that can create computer-interpretable guidelines, i.e. guidelines that a computer can run automatically. These languages, sometimes known as guideline representation models, include PROforma [2], Asbru [3], Arden Syntax [4], GLIF [5], GUIDE [6], SAGE [7] and others. The Arden Syntax was a pioneering rule-based effort and is perhaps one of the best known guideline representation models, while being famous for its ‘curly braces problem’: the lack of standardised patient data formats caused by the Arden Syntax’s local data definitions within Medical Logic Modules [8]

Objectives

Methods

Results

Discussion

Conclusion